Influence of Semiconductor Thickness and Molecular Weight on the

Mar 17, 2015 - Physics Department, The University of Jordan, Amman 11942, Jordan ... substantial μ's of 0.01–0.02 and ∼10–4 cm2 V–1 s–1, re...
0 downloads 0 Views 4MB Size
Forum Article www.acsami.org

Influence of Semiconductor Thickness and Molecular Weight on the Charge Transport of a Naphthalenediimide-Based Copolymer in Thin-Film Transistors Yevhen Karpov,† Wei Zhao,‡ Ivan Raguzin,† Tetyana Beryozkina,§ Vasiliy Bakulev,§ Mahmoud Al-Hussein,⊥ Liane Haü ßler,† Manfred Stamm,†,∥ Brigitte Voit,†,∥ Antonio Facchetti,*,‡,# Roman Tkachov,*,† and Anton Kiriy*,† †

Leibniz-Institut für Polymerforschung Dresden e.V., Hohe Straße 6, 01069 Dresden, Germany Polyera Corporation, Skokie, Illinois 60077, United States § Ural Federal University, Mira Street 28, 620002, Yekaterinburg, Russia ∥ Center for Advancing Electronics Dresden (CFAED), Technische Universität Dresden, 01062 Dresden, Germany ⊥ Physics Department, The University of Jordan, Amman 11942, Jordan # Center of Excellence for Advanced Materials Research (CEAMR), King Abdulaziz University, P.O. Box 80203, Jeddah, Saudi Arabia 21589 ‡

S Supporting Information *

ABSTRACT: The N-type semiconducting polymer, P(NDI2OD-T2), with different molecular weights (MW = 23, 72, and 250 kg/mol) was used for the fabrication of field-effect transistors (FETs) with different semiconductor layer thicknesses. FETs with semiconductor layer thicknesses from ∼15 to 50 nm exhibit similar electron mobilities (μ’s) of 0.2− 0.45 cm2 V−1 s−1. Reduction of the active film thickness led to decreased μ values; however, FETs with ∼2 and ∼5 nm thick P(NDI2OD-T2) films still exhibit substantial μ’s of 0.01−0.02 and ∼10−4 cm2 V−1 s−1, respectively. Interestingly, the lowest molecular weight sample (P-23, MW ≈ 23 kg/mol, polydispersity index (PDI) = 1.9) exhibited higher μ than the highest molecular weight sample (P-250, MW ≈ 250 kg/mol, PDI = 2.3) measured for thicker devices (15−50 nm). This is rather unusual behavior because typically charge carrier mobility increases with MW where improved grain-to-grain connectivity usually enhances transport events. We attribute this result to the high crystallinity of the lowest MW sample, as confirmed by differential scanning calorimetry and X-ray diffraction studies, which may (over)compensate for other effects. KEYWORDS: semiconducting polymer, thin-film transistor, electron mobility, morphology, crystallinity



INTRODUCTION

reduces material consumption. In addition, ultrathin transistors are of fundamental interest because, for the bottom-gate configuration, they enable direct probing of charge transport and film morphology with the aid of surface-sensitive techniques. Furthermore, high-quality ultrathin semiconductor layers may find applications in the fabrication of sensors and optically transparent devices. Although transistors utilizing small molecules have been successfully downscaled into one-molecule-thick (monolayer) semiconductor channels,14−17 fabrication of polymeric FETs with only a few molecular layers is far more challenging, and only a few examples have been demonstrated.18,19 In general,

Semiconducting π-conjugated polymers have attracted considerable attention in recent years as an important class of materials for applications in large-area electronic devices, such as polymeric field-effect transistors (FETs).1−6 Recent developments in material chemistry, physics, and device engineering have led to significant advances in FET performance and impressive mobilities for both holes (up to 14 cm2 V−1 s−1)7−9 and electrons (up to 6 cm2 V−1 s−1).10−12 To achieve such high mobility values, relatively thick semiconductor films in the 30− 80 nm range are normally used. It is known however that in an FET charge carriers are confined by the gate field in a few nanometer-thick accumulation layer formed at the semiconductor−gate dielectric interface,13 suggesting that the performance of ultrathin FETs may, in principle, approach that of much thicker devices. Ultrathin polymeric transistors would be of high technological interest for next-generation large-area electronics because using thin semiconductor films © XXXX American Chemical Society

Special Issue: Forum on Polymeric Nanostructures: Recent Advances toward Applications Received: November 7, 2014 Accepted: March 9, 2015

A

DOI: 10.1021/am507759u ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

Forum Article

ACS Applied Materials & Interfaces

plane mobilities that are observed in time-of-flight and diode measurements.37 The observations described above raise interesting questions as to whether spin-coating can be used to fabricate ultrathin P(NDI2OD-T2) films, the morphology of the film, and how FET devices would perform by progressively reducing the P(NDI2OD-T2) film thickness. In this work, we fabricated and evaluated the performance of FETs with semiconductor thicknesses systematically varied from ∼2 to 50 nm using three P(NDI2OD-T2) batches of different molecular weights. Two samples, indicated as P-23 with a MW ≈ 23 kg/mol and a polydispersity index (PDI) = 1.9 and P-250 with a MW ≈ 250 kg/mol and a PDI = 2.3, were synthesized by recently developed chain-growth polycondensation38−57 catalyzed by a palladium complex ligated by bulky and electron-rich tri-tert-butylphosphine.58 The commercially available P(NDI2OD-T2) with an intermediate molecular weight (MW = 72 kg/mol, PDI = 3.2), prepared by step-growth Stille polycondensation and designated here as P-72, was also investigated for comparison. The device performances were correlated to film morphologies studied by atomic force microscopy (AFM) and X-ray diffraction methods.

achieving highly performing FETs requires the presence of continuous pathways for charge carriers that extend throughout the whole semiconductor channel. This requirement is difficult to fulfill in ultrathin channel devices as this demands the formation of a large area two-dimensional (2D)-connected microstructure.20,21 Previously, small-molecule based semiconducting monolayer FETs were prepared by a self-assembled monolayer technique.17 A drawback of this process is that the self-assembly process is slow. Langmuir−Blodgett (LB)18 and Langmuir−Schäfer (LS)19 techniques were utilized for the fabrication of mono/multilayer FETs based on small-molecules and polymers, respectively; however, these methods may have limitations for scale-up. Polydiacetylene-based monolayer transistors were fabricated by solid-state polymerization of crystalline monomers; however, rather low FET performance was achieved in this case and this method lacks universality.22,23 In view of this, fabrication of polymeric monolayer transistors using high throughput techniques would be a technologically attractive solution given the possibility of using the minimal amount of material, which can further reduce fabrication costs. Following this approach, Zhang et al. recently developed a new deposition methodon-the-fly dispensing spin-coating which prevents unwanted dewetting of ultrathin semiconducting films on hydrophobic surfaces and allows the preparation of ultrathin films using a minimal amount of semiconductor materials.23 The bithiophene-naphthalene diimide copolymer (poly{[N,N′-bis(2-octyldodecyl)-naphthalene-1,4,5,8-bis(dicarboximide)-2,6-diyl]-alt-5,5′-(2,2′-bithiophene)} (P(NDI2OD-T2))10 is an n-type semiconductor.24−32 This polymer shows remarkably high electron mobility both parallel and perpendicular to the substrate plane and a relatively weak dependence of charge transport on processing conditions and gate insulator dielectric constant, which are uncommon for several semiconducting polymers.10,33−35 Furthermore, grazingincidence wide-angle X-ray scattering (GIWAXS) studies of ∼50 nm thick P(NDI2OD-T2) films fabricated by spin-coating revealed a bulk unconventional face-on orientation of P(NDI2OD-T2) molecules (with their π-stacking direction lying out-of-plane).34 Indeed, other highly performing polymers, such as polythiophenes, as well as donor−acceptor copolymers, such as diketopyrrolopyrrole-2a and benzothiadiazole-based8 copolymers, were found to exhibit a bulk edge-on orientation, supporting 2D charge transport in the π−π stacking.1−6 Following the discovery of P(NDI2OD-T2), several studies have attempted to rationalize charge transport in FETs and diode architectures based on this polymer34,35 as well as investigate details of the film’s morphology and how it varies with processing conditions and thermal history. A recent explanation for the efficient in-plane charge transport properties of P(NDI2OD-T2) thin films was proposed on the basis of near-edge X-ray absorption fine structure (NEXAFS) spectroscopy data.36 By comparing the bulk-sensitive and surfacesensitive NEXAFS data, McNeil et al. observed36 a molecular orientation at the surface of the film that was distinctly different from that of the bulk. Whereas a more “face-on” orientation of the conjugated backbone is observed in the bulk of the film, consistent with the lamella orientation observed by GIWAXS, a more “edge-on” orientation is observed at the surface of the film by surface-sensitive NEXAFS spectroscopy. This distinct edge-on surface orientation may also explain the high in-plane mobility that is achieved in top-gate P(NDI2OD-T2) FETs, whereas the bulk face-on texture accounts for the high out-of-



RESULTS AND DISCUSSIONS Charge Transport Measurements. To minimize processing variations, we fabricated the polymer films in a glovebox by spin-coating 0.5−8 g/L o-dichlorobenzene (DCB) and tetrachloroethane (C2H2Cl4) solutions, affording films with a nominal thickness of ∼2−50 nm (Table 1) as accessed by profilometry, optical absorption, and ellipsometry measurements (Figures S1 and S2 and Table S1 in the Supporting Information). The finished devices were then tested under ambient conditions. Representative transfer characteristics of these FETs are shown in Figure 1 and Figures S3 and S4 in the Table 1. FET Parameters for P-23, P-72, and P-250 Based Devices Fabricated in this Studya entry polymer 1 2 3 4 5 6 7 8 9 10 11 12 13

P-23 P-23 P-23 P-23 P-72 P-72 P-72 P-72 P-250 P-250 P-250 P-250 P-250

solvent/conc. (g/L)

hb (nm)

μelectron (cm2 V−1 s−1)

μhole (cm2 V−1 s−1)

C2H2Cl4/7 C2H2Cl4/3 C2H2Cl4/1 C2H2Cl4/0.5 DCB/8 C2H2Cl4/7 C2H2Cl4/3 C2H2Cl4/1 DCB/8 C2H2Cl4/7 C2H2Cl4/3 C2H2Cl4/1 C2H2Cl4/0.5

35−50c 16 5 2.5 50c 40c 15 5 50c 40c 14.5 4.5 2

0.30−0.46 0.31−0.37 ∼0.005 ∼10−4 0.20−0.30 0.27−0.33 0.27−0.28 0.01−0.015 0.15−0.31 0.10−0.30 0.10−0.20 0.01−0.02 ∼10−4

∼0.008 ∼0.001 ∼0.0003 ∼0.006 ∼0.009 ∼0.007 ∼0.001 ∼0.01 ∼0.009 ∼0.006 ∼0.001

a

W = 500 mm, L = 50 mm, Ci = 3.5−4.0 nF/cm2. The carrier mobility was measured in saturation. The mobility range indicates the minimum and maximum values, whereas the approximate sign indicates the order of magnitude values considering the low drain currents. Performance range tested for at least 10 devices. bFilm thickness on glass substrates as determined by optical absorption unless indicated (±10% accuracy). cThickness of the active layer was determined by profilometry (±5% accuracy). Note, the films deposited from C2H2Cl4 solutions are ∼3× thinner than those deposited from CHCl3 as determined by AFM, optical absorption, and ellipsometry data (see Supporting Information). B

DOI: 10.1021/am507759u ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

Forum Article

ACS Applied Materials & Interfaces

contrast to previous studies in which low-MW polymers were found to exhibit poorer performance than the corresponding high-MW ones.59−68 Transistors based on much thinner semiconductor films were also fabricated by spin-coating all of the polymer samples. Several polymer concentrations were used for the film deposition as indicated in Table 1, affording films from ∼2 to 15 nm. The electron mobilities of the ∼15 nm devices are significant with values of 0.31−0.37 cm2 V−1 s−1 for P-23 (entry 2), 0.27−0.28 cm2 V−1 s−1 for P-72 (entry 8), and 0.10−0.20 cm2 V−1 s−1 for P-250 (entry 11) (hole mobilities of ∼0.006− 0.008 cm2 V−1 s−1). Interestingly, these numbers are very similar to the electron/hole mobilities of the corresponding much thicker P(NDI2OD-T2) films. For transistors with even thinner active layers, both the electron and hole mobilities decline when approaching the nominal submonolayer thickness (Table 1, entries 3, 4, 8, 12, and 13). Thus, transistors with an ∼5 nm thick semiconductor film continue to exhibit substantial electron mobilities (0.02− 0.005 cm2 V−1 s−1, entries 3, 8, and 12); however, those based on the high MW samples perform 2−3 times better than the FETs based on the low MW samples. Remarkably, transistors with the active layer thickness of ∼2 nm, somewhat lower than the monolayer thickness (the thickness of edge-on oriented PNDIT2 molecules is ∼2.5 nm), showed a measurable electron mobility of ∼10−4 cm2 V−1 s−1 (Table 1, entry 13). However, further decreasing the semiconductor film thickness leads to full degradation of the FET performance. Interestingly, previously reported densely packed mono/multilayers of P(NDI2OD-T2) prepared by the LS technique with an exclusive edge-on orientation exhibited comparable electron mobilities.22,23 Thus, it is quite impressive that our ultrathin and poorly compacted films (see AFM discussion) can afford such efficient transport characteristics. Furthermore, our P(NDI2OD-T2) ultrathin transistors exhibit charge carrier mobilities that compete with or surpass those of previously reported ultrathin films fabricated by dip-coating.69 The influence of the molecular weight of P(NDI2OD-T2) on the performance of transistors with different semiconductor thicknesses merits further discussion. In many cases documented in the literature, conjugated polymers with very high molecular weights (assuming they remain soluble) possess superior charge-transport, morphological, and film-forming properties, and therefore show better performance in thin film transistors than their low molecular weight counterparts. For example, remarkable hole mobility up to 10 cm2 V−1 s−1 was recently reported for a high MW diketopyrrolopyrrolebased polymer, whereas the corresponding lower molecular weight polymers performed far more poorly.8 Similar trends were observed for polybenzothiadiazoles9 as well as other copolymers. From these precedents, the results of this work are somewhat surprising in view of the weak dependency of the carrier mobility on the polymer MW. To further investigate these trends, we carried out a detailed investigation of P(NDI2OD-T2) film morphology. P(NDI2OD-T2) Film Morphology. Atomic force microscopy (AFM) is a powerful tool for studying polymers at the thin-film and single molecule levels.70−72 Figure 2 shows AFM topography images of thin films prepared on silicon wafers by spin-coating solutions of P-250 and P-23 at concentrations varying from 0.01 to 1 g/L. Films prepared using 1 g/L (∼7 nm thick) had a uniform nanofibrous morphology (Figure 2a and i) similar to those previously observed for thin (25 nm)

Figure 1. Polymer structure, schematic representation of the FET, and representative n- and p-channel transfer characteristics measured under ambient conditions for P-250 FETs with the indicated thicknesses. VD = 100 and −100 V for n- and p-channel measurements, respectively. Gate leakage currents were omitted for clarity; however, they are ∼2−3 orders of magnitude lower than the drain currents at the maximum gate voltage for all well-behaving devices. Semiconductor film deposition details, except for the inactive ∼1 nm thick film, are shown in Table 1.

Supporting Information, and FET mobilities are collected in Table 1. FETs with the top-gate, bottom-contact configuration and 40−50 nm thick P-250 films fabricated on glass substrates with Au source/drain/gate electrodes and PMMA as the dielectric layer exhibit electron mobilities up to 0.31 cm2 V−1 s−1 (0.01 cm2 V−1 s−1 for holes) when measured under ambient conditions (Table 1, entries 9 and 10). The devices based on P72 (commercial N2200, entries 1 and 2) exhibit saturation mobilities of 0.2−0.3 cm2 V−1 s−1 for electrons and ∼0.006− 0.009 cm2 V−1 s−1 for holes (Table 1, entries 5 and 6), in agreement with previous studies.3,12 Surprisingly, to a small yet statistically significant extent, the low molecular weight sample P-23 exhibits larger electron mobilities than those of higher MW P-72 and P-250 samples for 40−50 nm thick semiconductor films. Thus, the P-23 devices show saturation mobilities of 0.30−0.46 cm2 V−1 s−1 for electrons and ∼0.008 cm2 V−1 s−1 for holes (Table 1, Entry 1). This result is in C

DOI: 10.1021/am507759u ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

Forum Article

ACS Applied Materials & Interfaces

Figure 2. Representative AFM topography (a−d, i, k) images and cross sections (e−h, j, l) of P-250 (a−h) and P-23 (i−l) films spin-coated at 2000 r−1 on Si wafers from 1 g/L (a, e, i, j), 0.1 g/L (b, f, k, l), and 0.01 g/L (c, d, g, h) solutions.

assumed that for ultrathin films a uniform 2D network as formed from the higher MW P-250 sample should exhibit better charge transport than the disconnected film morphology of the P-23 film. This data is fully consistent with the FET data for ultrathin (≤5 nm) films. On the basis of literature reports7−9 and taking into account general considerations about efficient charge percolation of 1D conductors, the high MW P-250 polymer sample should perform much better than the others, especially for ultrathin-based devices. However, despite the order of magnitude difference in MW, P-250 FETs performed only slightly better in the monolayer transistor configuration than those based on low MW P-23. Furthermore, for the thick film devices, P-23 displayed statistically higher mobility, which may be related to different crystallinities of the samples. To quantify film crystallinity, we carried out differential scanning calorimetry (DSC) and X-ray diffraction measurements for these two P(NDI2OD-T2) samples as discussed below. Thermal Analysis. The thermal properties of P-23 and P250 were investigated by DSC. Heating−cooling−heating cycles were performed to achieve a comparable thermal history of the samples. Both polymer samples are crystalline (Figure 3); however, P-23 shows a slightly higher crystallinity than P250, which follows from the melting enthalpy (Table 2). The much sharper melting peak of P-23 is due to the lower dispersity index of P-23 and its less entangled structure as a consequence of its smaller molecular contour length. Importantly, the lower MW sample has even higher melting and crystallization temperatures as well as a higher melting enthalpy than the high MW sample. These data indicate that the lower MW P-23 sample studied in this work has already reached the saturation regime of the melting temperature versus molecular weight dependence. As further evidence,

P(NDI2OD-T2) films prepared by dip-coating69 or thick (50− 100 nm) P(NDI2OD-T2) films fabricated by spin-coating (Figure S5 in the Supporting Information).10,37 Importantly, independent of the polymer molecular weight and solution concentration, the nanofibers in all of the samples exhibited essentially the same height (∼2.5 nm, Figure 2c, f, i, and j). This thickness value is close to the computed molecular size of the edge-on oriented N-(2-octildodecyl)-substituted naphthalene diimide unit and to the lamellar d-spacing found for P(NDI2OD-T2) films.69 The lateral width of the fibers was found to be ∼30−40 nm (Figure 2d and h). It should be noted that the height of the elementary fibrils would be much smaller (